verbal-ink-infographic

Machine vs. Human: Translators Put to the Test

This is a guest post by Verbal Ink.

In contemporary culture, we are used to computers being able to shock us. The idea of a computer doing something that a previous generation thought was impossible is, to us, the norm. That’s why what paradoxically shocks those of us well-versed in technology and the power of machines is the areas where they still fail to match us mere humans.

As an example, think about translation. Absorbing languages is a natural and easy task for young humans- a baby can learn to speak any language in the world. As we age, it becomes harder to learn new languages, and some people are better at it than others. However, it is usually a matter of memorization, commitment, and time investment. For that reason, once someone has learned more than one language, it is generally not a difficult task for them to translate words, ideas, and conversations from one language they know to another. Professional translators will tend to have a better grasp of the details of grammar, syntax, and structure, as well as larger vocabularies than most speakers, but even a novice will be able to translate rudimentary conversation and basic sentences.

On the other hand, machines have a great deal of trouble with translation. They are unable to really understand the ebb and flow of a conversation, understand the main idea of an argument, or understand how the audience of a translation should affect the word choice. While a computer translation program has a huge list of words in different languages, and therefore possesses a deeper and larger vocabulary than a human ever could, it is unable to understand what it is translating. As a result, while a machine might be able to assign some translated meaning to each individual word in a sentence, it is rarely able to produce coherent paragraphs or even sentences.

To test this theory, we tried two different, well, tests. First, we took some marketing material in Spanish. We fed it into Google Translate and asked a professional Spanish to English translator to translate it, and then compared the results. What we found confirmed our suspicions. While the human translator was able to deliver a message with accuracy, nuance, and style, Google was unable to deliver sentences that really made sense, and the translation was often overly literal.

In Round 2, we upped the ante. Starting from an audio recording of a speech in Spanish, we asked the human and the machine to first transcribe the spoken word into written Spanish and then translate the written Spanish into written English. Here, the human translation was even better. The human was able to hear and accommodate errors and peculiarities in the speaker’s delivery, transcribing them accurately, and preserving them in the translation. Google Translate simply didn’t know what to do with misspoken words, repetition, or a slight stammer. It was clear that the human translator was able to do more than just translate words- she was interpreting the meaning of the speaker and discern how his delivery differed from his intention.

So the next time you get worried about being replaced by a computer, don’t fret. There are still many tasks- like translation- where a computer is unable to come close to human performance.

verbal-ink-infographic

This is a guest post by Verbal Ink. Verbal Ink’s transcription services, translation services, and writing services make it easier for the customers to share their stories.

 


Translate pdf files Machine translation won’t replace professional translation. However, machine translator is an excellent tool for understanding the main message of a foreign language text. For example, Multilizer PDF Translator helps you to read foreign PDF documents.

5 thoughts on “Machine vs. Human: Translators Put to the Test”

  1. Very nice. I’ve been thinking about doing this with fiverr.com – create a passage and have it translated by ten of the many $5.00 translators for comparison purposes.

  2. Using Google Translate for this experiment isn’t really fair. There are way better MTs out there that can be tweaked to yield better results for each particular project. That said, even the most advanced piece of software cannot translate as good as a human translator, and it never will. Unless computer switch from the current system to something like quantum physics, they will never translate as good as humans.

  3. I agree with you that computer translations are not as good as human translations. This is because when we use the computer to do the translations, it will often translate the contents word by word only and not take into consideration the whole meaning of the contents. So some sentences will not make sense once it’s translated by computer.

    On the other hand, human translations are better because there are translators who can stop and think about the context of the contents that have to be translated before translating it. This makes this method of translated much more accurate.

Comments are closed.